QUALITATIVE AND QUANTITATIVE COMPARISON OF RENAL VS. HEPATIC ULTRASONOGRAPHIC INTENSITY IN HEALTHY DOGS

Introduction

Ultrasonography lacks specificity with regard to diseases of the renal parenchyma.^1–3 Although a relationship between renal cortical echogenicity and parenchymal disease has been demonstrated in people,^2,4–6 and it has proven useful in the detection of cystic disease,⁷ there is little or no correlation between ultrasonographic appearance of the renal parenchyma and specific diseases.^1,2,6,8,9 Increased renal cortical echogenicity in both dogs and humans is, in general, abnormal, but many diseases may be responsible.^4,8,10–12

Quantitative assessment of renal cortical echogenicity is not routinely performed clinically, and has only recently been described in humans.^13,14 It has been recommended as potentially useful in the early detection of some renal diseases, such as aminoglycoside-induced nephrotoxicity, and was reported to be more sensitive than visual inspection of ultrasound images early in the disease.¹⁵ In cats, renal cortical echogenicity varies in normal animals, being largely dependent on the degree of fat vacuolization in cells in the proximal tubular epithelium.^10,14,16 Qualitative evaluation, on the other hand, is subjective and variable. In addition, in both people and animals, changes in ultrasound reflectivity can be measured quantitatively while not being apparent visually.^15,17 For standardization, renal cortical echogenicity is often compared with that of the liver, spleen, renal sinus, and/or renal medulla.^{1,2,6,8,10,11,18–22}

Although approximately 10% of normal dogs may have hyperechoic renal cortices,¹⁹ there is general acceptance that normal renal cortical echogenicity in dogs should be either isoechoic or hypoechoic to liver.¹¹ This relationship has been derived from human studies, and although a few experimental studies exist, there are no careful comparisons of echogenicity of canine liver and kidney in healthy dogs.^15,23–25

Echogenicity of an organ is subjectively assessed by visual evaluation of the organ's brightness. Quantitatively, echogenicity should be measured through determination of the backscatter coefficient, but mean pixel intensity can be substituted. The aim of this study was to qualitatively and quantitatively examine the echogenicity of the renal cortex relative to the liver in normal, healthy dogs, using standard and harmonic imaging at 8.0 MHz. We hypothesized that the ultrasonographic signal intensity of the renal cortex would be higher than that of the liver.

Materials and Methods

Adult dogs owned by faculty, students, and staff at the Matthew J. Ryan Veterinary Hospital of the University of Pennsylvania were recruited for participation in this study. The inclusion criteria were: (1) healthy dogs 1–8 years of age, (2) no history of hepatic or renal disease, (3) no current clinical signs attributable to hepatic or renal disease, and (4) no intravenous fluid administration within the last 48 h.¹³ Subjects were fasted for ≥12 h, and positioned without the use of chemical restraint in left recumbency during the ultrasonographic evaluation. Owner consent was obtained before acquisition of data, and the protocol was approved by the University of Pennsylvania Institutional Animal Care and Use Committee.

Ultrasonographic examinations were performed with a GE Medical LOGIQ 9 Ultrasound Imaging System^* by one author (M.I.). Hair was clipped and acoustic gel applied to the skin. Each animal was imaged using an 8.0-MHz microconvex transducer. Using standard B-mode gray-scale ultrasonography, a sagittal image of the cranial pole of the right kidney adjacent to the caudate lobe of the liver was obtained using a right dorsal intercostal window. The gallbladder was avoided due to resultant distal acoustic enhancement within the hepatic parenchyma, and an acoustic window was selected that did not include shadowing from a rib or gastroduodenal gas. Three such images were obtained consecutively, repositioning the transducer each time, and the protocol was repeated using tissue harmonics mode to acquire three additional images. One focal zone was always used and was positioned at the level of the mid-right kidney. For each individual patient, the settings for gain, image depth, and focal zone depth were kept constant, but they varied between different subjects based on body size and image quality. The liver and both kidneys were closely evaluated for evidence of ultrasonographic abnormalities.

Following ultrasonographic evaluation, a blood sample and a urine sample (3 ml each) were collected by venipuncture and cystocentesis or free-catch, and submitted for routine analysis.

All images were qualitatively examined independently by two radiology residents who were unaware of the other's interpretation. For each group of three images, the right renal cortex was subjectively classified as hypoechoic, isoechoic, or hyperechoic to liver. The evaluation was conducted on both standard B-mode images and harmonic images.

Images were also analyzed using digital image analysis software.† For each image, an equally sized region of interest (ROI) was drawn within the renal cortex and adjacent liver at equal depths, and quantification of mean pixel intensity within each ROI was derived. A uniform depth selection was necessary to account for increased tissue attenuation of ultrasound with increased depth. The mean pixel intensity was determined using an 8-bit gray scale, with 256 shades of gray. Two series of measurements with different ROI geometries were obtained. First, the ROI was centered at the focal zone with inclusion of the largest possible area that could be applied to the renal cortex and a similarly sized area chosen in the liver (Fig. 1A and B). These ROIs had an elongated half-annular geometry, following the anatomic shape of the renal cortex in the cranial pole of the kidney. The long axis of the ROI was primarily oriented in the direction of the ultrasound beam; therefore, the mean pixel intensity values obtained were the result of a combination of backscatter (echogenicity) and attenuation. Second, a smaller square ROI was placed as close as possible to the near field within the kidney and the liver (Fig. 1A and C). The smaller square ROI position was designed to derive a measurement depending mostly on backscatter with minimal influence from attenuation, to reflect more closely the echogenicity of both organs.

Care was taken not to include any artifact or other tissue within an ROI, including renal medullary parenchyma, renal capsule, corticomedullary rim, and distal acoustic or edge shadowing. For each ROI geometry, the measurements from each subject were averaged to determine the individual renal and hepatic mean pixel intensities for each subject. This was repeated with images captured using the harmonic settings.

Results are presented as mean with a 95% confidence interval (CI). Values obtained during quantitative analysis were compared between groups using a two-sample Wilcoxon's rank-sum test. κ statistics were used to measure interobserver agreement for the qualitative analysis. The percentages of dogs having a renal cortex qualitatively hyper-, iso-, or hypoechoic to the liver were compared with published percentages observed in a group of healthy geriatric dogs imaged at 4–5 MHz using a χ² goodness-of-fit test (Table 1).¹⁹ Statistical significance was defined as P<0.05. All statistical analyses were performed using a commercial statistical software package.‡

Results

Forty-one dogs were initially enrolled. Sixteen were excluded based on abnormal biochemical or urinalysis data, abnormal ultrasonographic appearance of the liver or kidneys, or the inability to obtain diagnostic images due to patient size, lack of patient cooperation, or presence of gastrointestinal gas. Of the 25 remaining dogs, there were 18 neutered females and seven neutered males. The mean age was 3.2 years (1.3–8.0 years) and the mean weight was 16.2 kg (2.1–42.9 kg). In addition to 13 mixed-breed dogs, breeds represented included one of each of the following: Cockapoo, Weimaraner, Miniature Dachschund, Yorkshire Terrier, Pug, German Shepherd Dog, Whippet, Italian Greyhound, English Springer Spaniel, English Cocker Spaniel, Beagle, and Golden Retriever.

The renal cortex was qualitatively assessed as hyperechoic to the liver in 13/25 dogs (52%) on standard images and in 14/25 dogs (56%) on harmonic images. The renal cortex was qualitatively assessed as isoechoic to the liver in 10/25 dogs (40%) on standard images and in 11/25 dogs (44%) on harmonic images. There was fair agreement between the two observers for qualitative evaluation of standard (κ=0.43, P=0.005) and harmonic images (κ=0.51, P=0.005). Our qualitative assessment results were different than those reported in geriatric healthy dogs (P<0.001).¹⁹

Half-Annular ROI

With B-mode imaging, the mean pixel intensity of the renal cortex was 60 (95% CI 55–65) vs. 48 (95% CI 44–52) for liver. With harmonic imaging, the mean pixel intensity of renal cortex was 72 (95% CI 65–79) vs. 56 (95% CI 50–63) for liver (Fig. 2A). On B-mode images, the mean pixel intensity of the renal cortex was higher than in liver in 19/25 dogs (76%), equal to liver in 4/25 dogs (16%), and lower than liver in 2/25 dogs (8%). On harmonic images, the mean pixel intensity of the renal cortex was higher than in liver in 22/25 dogs (88%), equal to liver in 2/25 dogs (8%), and lower than the liver in 1/25 dogs (4%). The mean pixel intensity of renal cortex was significantly higher than in liver, for both standard (P=0.0007) and harmonic (P=0.0107) tissue imaging. In addition, the mean pixel intensities measured in the right kidney were significantly higher with harmonic imaging than with B-mode imaging (P=0.0220). There was no statistically significant difference between the mean pixel intensity of liver measured on standard vs. harmonic imaging for the liver (P=0.0535). The mean differences in mean pixel intensity between liver and kidney was higher with harmonic imaging (mean=15; 95% CI 10–21) than in standard gray scale (mean=12; 95% CI 8–16), although the difference was not statistically significant (P=0.5).

Discussion

Ultrasonographers regularly use subjective assessment of relative organ echogenicity. The visual perception of echogenicity is a result of ultrasonic backscatter and attenuation.^17,23–25 Techniques are available to measure backscatter and attenuation separately, but these techniques were not available to us.^17,24 As perceived echogenicity is what is routinely used clinically to discriminate between normal and abnormal during an ultrasound examination, we decided to use mean pixel intensity as a simple way to quantify the visual impression of the ultrasonographer. We also defined two ROI geometries, one that assessed the effect of both backscatter and attenuation (half-annular ROI) and the other that minimized the effect of attenuation by using near-field positioning and a small ROI surface area. Although the difference was not great, the renal cortex did have significantly higher mean pixel intensity than liver with the half-annular ROI that took both attenuation and backscatter into account. The renal cortex was found to be hyperechoic to adjacent liver with the half-annular ROI geometry in 76% and 88% of normal dogs when using standard and harmonic imaging, respectively.

Clinically, it is the subjective assessment of echogenicity that is used to make decisions. Our qualitative evaluation was in agreement with the quantitative data, with renal cortex having higher mean pixel intensity than liver in 52% of dogs on conventional gray-scale imaging and in 56% of dogs on harmonic imaging. This is in contrast to prior work where the kidney was hyperechoic to liver in 3% of dogs, isoechoic to liver in 36% of dogs, and hypoechoic to liver in 61% of dogs.¹⁹ In that study geriatric dogs were used and scanning was at 4–5 MHz, which might explain the difference. Therefore, although we do not suggest that quantitative measurement of echogenicity should be carried out routinely, when imaging dogs at 8.0 MHz, which is a commonly used imaging frequency, the kidney can be slightly hyperechoic to the adjacent liver in the absence of hepatic or renal disease.

We found an even greater difference in renal cortical vs. hepatic echogenicity when harmonic imaging was used. The difference in mean pixel intensity was 16 on harmonic images vs. 13 on conventional gray-scale images. In humans, some tissues can look artificially hyperechoic when harmonic imaging is used, leading to erroneous interpretation.²⁶ For each organ in our study, the mean pixel intensities were higher using harmonic images compared with conventional gray-scale imaging, and the difference was statistically significant for the kidney using the half-annular ROI and for both kidney and liver using the superficial square ROI. This finding highlights the importance of recognizing how inherent normal tissue echogenicities differ when using fundamental frequency ultrasonography vs. image processing techniques, such as harmonics.

As previously stated, image depth, focal zone position, and gain were not uniform between dogs, and all of these factors affect echogenicity. However, the variability in size of the dogs precluded keeping these variables constant. In addition, although overall echogenicity is dependent on these factors, the relative difference in echogenicity between liver and kidney should have been unaffected in each image.

The 8.0-MHz frequency used in our study is higher than the 3.5–5.0 MHz used in humans and 4.0–7.5 MHz used in small animals where renal cortical echogenicity was found to be isoechoic or hypoechoic to liver.^{6,8,11,13–15,19,21} Transducer frequency is known to affect echogenicity,³ and the effect of frequency on the observed echogenicity has been recognized in canine kidney^19,27 and human fetal bowel.^26,28 The frequency dependence of tissue echogenicity is in part due to backscatter from Rayleigh scattering within tissue.^29,30 When ultrasound frequency increases, more components of the tissue can become Rayleigh scatterers, and these generate echoes whose intensities are proportional to the fourth power of the frequency.³¹ For example, in the 1–5-MHz range, the echo from blood is approximately 20 dB below that from tissue. In the 20–30-MHz range, however, blood cells become Rayleigh scatterers, and blood becomes isoechoic to tissue.^31,32 Therefore, the higher frequency we used may in part explain our results, if relatively more Rayleigh scatterers are present in the renal cortex than in the liver at higher frequencies.

Attenuation of ultrasound is also known to be directly dependent on ultrasound frequency. As we observed more differences in mean pixel intensities using the deep half-annular ROI (attenuation and backscatter) than the superficial, small square ROI (backscatter mostly), it can be hypothesized that the overall difference in mean pixel intensities across the imaged regions of kidney and liver is mostly due to the fact that the liver is more attenuating than the renal cortex. In addition, cortical anisotropy of the kidney could have contributed to the differences observed using the deep half-annular ROIs; indeed, this directionally dependent hyperechogenicity where cortical medullary rays are perpendicular to the beam is expected to be maximized within the area of the cranial pole of the kidney when imaged in long axis.³³ Regardless of the underlying mechanism, we found agreement between the differences in mean pixel intensity measured using the half-annular ROI and visual assessment of the relative overall echogenicity. Thus, the renal cortex being more often hyperechoic to the liver likely represents the norm at 8.0 MHz.

Histopathologic examination of kidney and liver was not performed, and dogs were classified as healthy based on history, lack of clinical signs, ultrasonographic examination, and laboratory data. Although histopathologic data would have been the gold standard, it was clearly not possible to obtain in this study. In addition, while normal hematologic parameters do not entirely rule-out diseases or benign fatty changes, we believe that the consistency of the results provides reasonable evidence that the dogs were free of hepatic or renal disease.

Another limitation was that no analysis was carried out on the left kidney. This occurred for two reasons: (1) the lack of a comparable adjacent internal standard such as the spleen,¹¹ and (2) the inability to maintain constant settings for depth, gain, and focal zone position between the two kidneys, thereby preventing quantitative comparison between the right and left kidney. We focused on addressing the currently accepted specific relationship between the echogenicity of the right kidney and adjacent liver.^11,19,21 Being that this relationship is commonly evaluated during routine imaging, we believe our findings are of clinical relevance.

Future studies addressing our hypothesis with a larger sample size would be beneficial, as would the study subjects for which histopathologic renal and hepatic analysis could be performed. In addition, the effects of varying the number and depth of focal zones could be explored, as well as the significance of using a variety of ultrasound transducers and frequencies.

In conclusion, we found that the renal cortex imaged sonographically at 8.0 MHz can be mildly hyperechoic relative to the liver in normal dogs. These findings redefine the framework within which the renal cortical parenchyma is routinely evaluated, and suggest that mild hyperechogenicity of the renal cortex in comparison with the liver is a normal finding in some dogs.